How to Configure a Custom LLM Proxy in OpenOPC (and Debug LiteLLM) A quick walkthrough on fixing LiteLLM routing errors when hooking up custom OpenAI-compatible proxies in OpenOPC. Dev.to · 4小时前
I wrote down 61 rules, then failed my own repos Nobody starts a library because they wanted to write a SECURITY.md. You start it because there is a... Dev.to · 12小时前
Designing MCP Tools for a 7B Model, Not a 70B One I built an agentic assistant for battery engineering. A physics-based digital twin runs the actual... Dev.to · 13小时前
Your MCP server's real constraint is the context window, not the API A local stdio MCP server cannot serve claude.ai. Building the hosted version took a day; working out what it should refuse to do took the rest. The token arithmetic, the excerpt scanner, and the four Dev.to · 13小时前
How Do You Build an Evaluation Harness for AI Agents? You have an agent that works. Now someone asks how you know, and the honest answer is that you tried... Dev.to · 15小时前
Your model doesn't need to pass the bar exam. It needs to parse a log file. Intro Every few weeks, another frontier model claims a new benchmark, better reasoning,... Dev.to · 15小时前
You Forgot to Mention the 500 Million Rows The query is genuinely elegant. It's also a self-join across a table you never mentioned has half a billion rows. It ran fine against your dev sample. Dev.to · 17小时前
Your AI agent can't design images. It can write HTML. Diffusion models can't do layout, but your agent already writes HTML. MCP setup for Claude Code and Cursor, plus the self-review loop that makes it work. Dev.to · 17小时前
I gave a user the right answer with the wrong mechanism. They caught it within a day. Last week I wrote about saying no to a feature request: a user designed a laneId mode for my project... Dev.to · 18小时前
Spring AI Prompt Caching and Chat Memory: Where the Tokens Go — LLM Cost Control 2/4 Suppose the metrics are in place, each task has the model it actually needs, and every feature has... Dev.to · 19小时前
Understanding Over Origin: The Missing Friction A few days ago, I wrote "Understanding Over Origin" and it got alot of engagement and I'm really... Dev.to · 22小时前
Hardening Claude Code: the Minimum Floor as Config Ten Claude Code settings that survive an agent trying to route around them, each paired to the exact bypass it closes, verified against live docs. Dev.to · 22小时前
LLMs on Consumer Hardware — Part 2: Prefill and the Failure of the AI PC The second entry in a build-log series on local LLMs: the two phases of inference, a controlled cross-machine benchmark, the disk cost of loading a model, and a comparison against a free-tier cloud mo Dev.to · 22小时前
I built a tool that roasts your code with regex — no AI involved The problem In 2026, devs spend 11.4 hours a week reviewing AI-generated code — more time than they... Dev.to · 23小时前
You Can't Unit-Test an LLM. Here's What I Built Instead. Every team shipping an LLM feature eventually hits the same wall: the thing you built is... Dev.to · 23小时前
